HEALTH AND HEALTH BEHAVIORS DURING THE WORST OF …According to data from the National Health Expenditure Accounts, the health care spending represents a significant and increasing

NBER WORKING PAPER SERIES

HEALTH AND HEALTH BEHAVIORS DURING THE WORST OF TIMES

Erdal TekinChandler McClellanKaren Jean Minyard

Working Paper 19234http://www.nber.org/papers/w19234

NATIONAL BUREAU OF ECONOMIC RESEARCH1050 Massachusetts Avenue

Cambridge, MA 02138July 2013

Previously circulated as "Health and Health Behaviors during the Worst of Times: Evidence from the Great Recession." Erdal Tekin gratefully acknowledges support from the Gary and Stacey Jacobs Fellowship. The views expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research.

NBER working papers are circulated for discussion and comment purposes. They have not been peer-reviewed or been subject to the review by the NBER Board of Directors that accompanies official NBER publications.

© 2013 by Erdal Tekin, Chandler McClellan, and Karen Jean Minyard. All rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided that full credit, including © notice, is given to the source.

Health and Health Behaviors during the Worst of Times Erdal Tekin, Chandler McClellan, and Karen Jean Minyard NBER Working Paper No. 19234July 2013, Revised January 2017JEL No. E32,I00,I10,I12,I14,I15

ABSTRACT

Previous studies have shown that recessions are typically associated with better health and health behaviors. However, majority of these studies use data from the periods of relatively milder economic downturns that predate the “Great Recession.” In this paper, we examine the relationship between measures of macroeconomic conditions and a large set of outcomes of health and health behaviors using data from BRFSS between 1990 and 2014 with an emphasis on the period enveloping the Great Recession. Our results provide some support for the notion that weaker macroeconomic conditions are positively associated with health related outcomes, although the evidence is stronger for some of the outcomes (e.g., smoking and physical exercise) than others and is not present for some of the other outcomes (e.g., experiencing poor mental health) at all. But overall, the estimates are too small and imprecisely estimated to have any meaningful implications. Furthermore, the estimates become weaker over time and largely disappear in years prior to the period of the Great Recession and continues to remain small during periods enveloping the recession and the afterwards. Overall, we found no evidence that the Great Recession had a significant influence on the existing trends in health and health behaviors.

Erdal TekinSchool of Public AffairsAmerican University4400 Massachusetts Avenue NWWashington, DC 20016-8070and IZAand also [email protected]

Chandler McClellanCenter for Behavioral Health Statistics and Quality (CBHSQ):Substance Abuse and Mental Health Services [email protected]

Karen Jean MinyardDepartment of Public Management and Policy and Georgia Health Policy CenterAndrew Young Policy StudiesGeorgia State UniversityP.O. Box 3992Atlanta, GA [email protected]

3""

I. Introduction

The most recent recession in the United States, commonly referred to as the “Great

Recession,” differs significantly from any other economic crisis since World War II, and closely

resembles the Great Depression, particularly in terms of its severity and duration. The recession,

which lasted for 18 months, officially ended in June 2009, but the recovery had been slow,

particularly in job creation with the unemployment rate remaining at a stubbornly high level until

recently. According to the Bureau of Labor Statistics, between 2007 and 2010, the average

unemployment rate more than doubled from 4.6 percent to 9.6 percent. During the same period,

the employment rate decreased from 63 percent to 58.5 percent and approximately 5.2 million

individuals left the labor force.1

The fact that the recent economic downturn was both deeper and longer than the previous

recessions prompts an important question: to what extent has the Great Recession affected health

and health behaviors of Americans, and has the rising unemployment or decreasing employment

experienced during the Great Recession had a different effect on these outcomes than in previous

recessions? Although there is an extensive literature on the relationship between economic

conditions and health outcomes, the coverage period of the majority of existing studies predate

the period of Great Recession. But it is not clear whether the findings from these studies can

serve as a reliable guide to the relationship in more recent years for several reasons. First, the

sharp rise in unemployment and poor prospects of re-employment experienced during the Great

Recession may have led to an unprecedented number of Americans struggling with financial

strain. Consequently, the poverty rate rose sharply from 12.5 percent to 15 percent between 2007

and 2011 (U.S. Census Bureau 2012). Furthermore, the caseload for the Supplemental Nutrition

1 See http://www.bls.gov/cps/cpsaat01.pdf.

4""

Assistance Program (SNAP) reached record levels, with about one in seven Americans

participating in the program in 2011 (U.S. Department of Agriculture 2012). Second, the

collapse of the United States housing market in the same period and the subsequent decline in the

housing prices caused foreclosure rates to rise to historically high levels, further creating a

difficult burden on the lives of millions of Americans. As a culmination of a period of financial

strain, foreclosure often represents both a highly stressful event and a major shock to family

wealth, effects of which might be exacerbated by high unemployment rates during a recession

(Currie and Tekin 2015). Third, there was a sharp decline in the proportion of the population

with employer-sponsored health insurance during the Great Recession, driven by the significant

loss of employment during that time (e.g., Cawley et al. 2011, Holahan 2011, White and

Reschovsky 2012). According to Cawley et al. (2011), the number of Americans who lost health

insurance during the Great Recession was about nine times higher than the number who lost

insurance during the previous [2001] recession. Finally, the recession took a particularly

significant toll on local governments and their health departments across the United States,

undermining their capacity to provide critical core public health services for local communities

(Willard et al. 2012). Therefore, it is not clear whether one can make inferences from the studies

of earlier periods that would reliably apply to the most recession and the years since then.

This paper provides a fresh look at the relationship between macroeconomic conditions

and health and health behaviors. We use individual level data from the Behavioral Risk Factor

Surveillance System (BRFSS) to examine the relationship between macroeconomic conditions

and individuals’ health risk behaviors between 1990 and 2014. The range of outcome measures

we consider include modifiable health behaviors pertaining to smoking, alcohol consumption,

and physical activity, as well as self-reported general physical and mental health. While the

5""

BRFSS was used previously to study the relationship between unemployment and health

outcomes of individuals, our study is novel in terms of studying a large set of outcomes

reflecting both health and health behaviors for a period that leads up to, coincides with, and

follows the Great Recession. Furthermore, the previous studies usually rely on aggregate

unemployment rate as the proxy for the business cycle.2 In our analysis, we provide evidence

from two measures to capture macroeconomic conditions. The first measure is state

unemployment as utilized by the extant literature. The second measure that we use is the

employment rate, measured by the percentage of the state population who has a job. This

measure is potentially important because one of the key developments during the Great

Recession and the period that follows was the sharp rise in the number of people who have left

the labor force after losing hope of finding a job. Therefore, an analysis relying solely on

changes in the unemployment rate may not properly reflect a state’s economic condition. For

example, an improvement in the economic outlook may cause discouraged workers who had left

the labor force earlier to re-enter, leading to an increase in the unemployment rate or a prolonged

period of high unemployment. Conversely, a decrease in unemployment may reflect that

individuals are leaving the labor market after giving up hope of finding a job, rather than

improving economic conditions. However, despite these points it turns out that our results are

extremely consistent between these two measures.!

Our analysis also has the potential to provide useful information from a policy

perspective. According to data from the National Health Expenditure Accounts, the health care

spending represents a significant and increasing burden on the U.S. economy, accounting for

17.5 percent of the nation’s annual gross domestic product (Centers for Medicare & Medicaid

2 Several exceptions to this include Ruhm (1995; 2000; 2005), who employ measures of employment rate.

6""

Services, 2016). If, in fact, the sharp increase in unemployment during the Great Recession had

an adverse effect on health and, therefore is partially responsible for the rise in health care costs,

then this information could provide further basis for government interventions to reduce the

unemployment rate or mitigate its effects, including the American Jobs Act and expansions in

the Unemployment Insurance Program.

II. Literature

There is a sizeable literature examining the relationship between macroeconomic

fluctuations and health. In a widely cited study, Ruhm (2000) found that mortality and morbidity

in the United States follow a pro-cyclical pattern, i.e., they both worsen when the economy

temporarily improves. This finding was largely confirmed by a series of follow-up studies by

Ruhm (e.g., Ruhm 2003, Ruhm 2005, Ruhm 2007) and was echoed by others as well (e.g.,

Dehejia and Lleras-Muney 2004). To a large extent, a similar pattern was also established for

other countries (e.g., Neumayer 2004, Granados 2005, Gerdtham and Ruhm 2006, Ásgeirsdóttir

et al. 2014).!

Motivated by a desire to explain the mechanisms for the pro-cyclicality in mortality, this

finding has spawned a new wave of research focusing on the relationship between

macroeconomic conditions and health behaviors. The overall finding from these studies is that

health behaviors mostly improve during economic downturns. One common explanation offered

for this finding was that recessions lead to changes in certain life styles, which are health

promoting. In particular, it was argued that work itself is stressful and a reduction in time at work

may reduce the prevalence of stress-induced illness as well as make more time available for

salutary behaviors, such as exercise and healthy dieting (e.g., Catalano, Goldman-Mellor et al.

2011). For example, Ruhm (2005) used data from the BRFSS between 1987 and 2000 and found

7""

that changes in health behaviors represent a key mechanism for the pro-cyclical variation in

mortality and morbidity observed in the literature. In particular, he showed that smoking and

excess weight decline during temporary economic downturns, while leisure-time and physical

activity increase. In another study, Ruhm and Black (2002) used data from 1987 to 1999 waves

of the BRFSS to show that alcohol consumption decreased in bad economic times. The authors

concluded that any stress-induced increases in alcohol consumption during recessions were more

than offset by income effects.3

More recently, Xu (2013) combined health data from the BRFSS between 1984 and 2005

and the National Health Interview Survey (NHIS) between 1976 and 2001 along with

employment data from the Current Population Survey (CPS) to examine the relationship between

wages and hours of work on health behaviors of low-educated individuals. He found that higher

wages and hours of work tied to economic expansions were associated with increased smoking

and decreased physical activity. His study also provided evidence suggesting that economic

downturns may have heterogeneous effects on health behaviors based on the extent to which

behaviors are time-intensive.

Studies using data sources other than the BRFSS usually reach similar conclusions,

although the consensus appears to be less robust and the findings to be more sensitive to the time

period considered. For example, Ettner (1997) used data from the 1988 National Health

Interview Survey (NHIS) and found that non-employment significantly reduced both alcohol

consumption and dependence symptoms, possibly due to an income effect. Ásgeirsdóttir et al.

(2014) examined the effect of the October 2008 economic crisis in Iceland on a range of health

3 In contrast, Dee (2001) finds that the prevalence of binge drinking is strongly counter-cyclical using data from the BRFSS over the period of 1984-1995. Ruhm and Black (2002) offer a number of explanations for the contradictory finding in Dee (2001), including the relatively small number of states contained in the BRFSS in early years and the lack of sampling weights in the analysis performed in Dee (2001).

8""

behaviors and found that the crisis led to reductions in health-compromising behaviors, including

smoking, heavy drinking, consumption of sugared soft drinks and fast food, and indoor tanning.

Furthermore, they documented that the crisis reduced consumption of fruits and vegetables, but

increased consumption of fish oil and getting recommended hours of sleep. The authors also

argued that these behavioral changes were explained by increases in prices during the crisis. On

the other hand, Böckerman et al. (2007) explored the relationship between weight and economic

conditions using Finnish data from the period of 1978-2002. They found that an improvement in

economic conditions, measured by the regional unemployment rates, resulted in a decrease in

Body Mass Index (BMI). Similarly, Charles and DeCicca (2008) used data from the NHIS for

the years 1997-2001 to document evidence of a pro-cyclical relationship for weight-related

health and mental health among men.

Most recently, several studies documented that the pro-cyclical relationship between

macroeconomic conditions and mortality obtained in earlier studies might be weakening. For

example, Ruhm (2015a) revisited the aggregate mortality and unemployment relationship using

data from multiple sources over the period between 1976 and 2010. He found that the strong

procyclical pattern of mortality present in the 1970s and 1980s has largely disappeared or even

become countercyclical in recent years. One exception to this is the deaths from cardiovascular

disease and transport accidents. Interestingly, Ruhm (2015a) showed that the decline in the

procyclicality of mortality began long before the Great Recession and in fact increased slightly in

the recession. In another recent study, Ruhm (2015b) extended the analysis period to 2013 and

confirmed the earlier findings of a declining procyclical fluctuation in U.S. mortality.

Furthermore, this analysis also revealed that the two severe economic crises occurring at the

beginning of1980s and the Great Recession both had a protective effect on mortality that was

9""

about twice as large as that estimated by the higher unemployment rates experienced during such

periods alone.

Relatedly, McInerney and Mellor (2012) examined the relationship between recessions

and seniors’ health and health behaviors using data from the Medicare Beneficiary Survey. They

found that the relationship between unemployment and mortality of the elderly was negative for

the analysis period of 1976-1991, but positive for the period of 1994-2008.

It is important to stress that the existing evidence was largely built on information that

predates the period of the Great Recession, with the exception of the two recent Ruhm studies

mentioned above and a handful other studies (e.g., Macy et al. 2013; Colman and Dave 2013,

2014). Macy et al. (2013) used data from a county in Indiana to study the health behaviors of

individuals between 2005 and 2011. The authors found that financial strain reported during the

great recession had a negative impact on a range of health behaviors overall. Colman and Dave

(2013) examined the relationship between unemployment and physical activity using data from

the American Time Use Survey. Using within-state variation in employment and unemployment,

the authors found that recreational exercise increased during recessions as employment

decreased. In addition, they also found that individuals substituted into television watching,

sleeping, childcare, and housework. In a follow up study, the same authors studied the effects of

individual experience of unemployment on a range of health behaviors during the most recent

recession using data from the Panel Study of Income Dynamics and the National Longitudinal

Survey of Youth 1979 (Colman and Dave 2014). Their findings were largely consistent with

prior studies based on cross-sectional data, i.e. becoming unemployed was associated with a

small increase in leisure-time exercise and in body weight, a moderate decrease in smoking, and

a substantial decline in total physical activity.

10""

III. Data

The Centers for Disease Control and Prevention’s Behavioral Risk Factor Surveillance

System (BRFSS) is an on-going health survey system tracking the health conditions and risk

behaviors in the United States since 1984. Steadily expanding from 15 states in 1984, the BRFSS

is a representative telephone survey that currently provides coverage of all 50 states and the

District of Columbia. Over the course of each year, the BRFSS contacts over 200,000 individuals

to create a repeated annual cross section. The present study focuses on the period from 1990 to

2014, for which the sample size totals around 2.36 million individuals after limiting the analysis

to individuals ages 25 to 55 and excluding missing observations on key variables.4 Descriptive

statistics for the health behaviors and all of our covariates are presented in Table 1.

Measures of Macroeconomic Conditions

Our first measure of economic conditions is the state unemployment rate. This measure is

obtained from the Local Area Unemployment Statistics (LAUS) of the Bureau of Labor Statistics

(BLS). As shown in Table 1, the average state unemployment rate was 6.3 percent during our

analysis period. We take the average of this measure over a three-month period because

contemporaneous values might not reflect the true economic conditions in a state, but rather

capture short-term fluctuations.5 This is similar in principle to the approach in studies using

annual data, which basically smooth over a 12-month period.

Our second measure is based on the percentage of individuals employed at the state level,

again averaged over a three-month period. In particular, we construct the average percentage of

the civilian non-institutionalized state population (aged 16 and over) employed during the three

4 Note however that the sample size for each regression slightly differs from the baseline sample size due to the number of observations in the outcome variables. 5 Note that we also estimate our models using one-month-lagged measures of these variables. These results produced estimates that are very similar to those presented in this paper and are available from the authors upon request.

11""

months ending with the survey month.6 This measure is also available from the BLS’s LAUS.

As illustrated in Table 1, the average state employment rate was 61.3 percent during our analysis

period.

Outcome Variables

Topics included in BRFSS provide an extensive overview of a respondent's current

health, health history, and health behaviors. Current health questions range from broad ones,

such as those asking about general health, to specific ones, such as questions asking if the

respondent snores. Included in this range are questions regarding smoking and drinking behavior,

stress, and mental health. In addition to detailed health questions, the BRFSS provides

information on the typical demographic characteristics of its respondents.

We focus on smoking behavior as the first domain of outcomes in our study. The

“Current Smoker” outcome is an indicator variable equal to 1 if the respondent is a current

smoker, and 0 otherwise. BRFSS provides information on the smoking behavior based on

whether the respondent smokes daily, some days, is a former smoker, or has never smoked. The

current smoker variable indicates whether the respondent smokes daily or some days. Similarly,

“Daily Smoker” indicates whether the respondent smokes every day. In the past, smoking has

typically been considered a normal good and has exhibited a pro-cyclical pattern (Bobak, Jha et

al. 2000, Ruhm 2000). However, the prevalence of tobacco use has changed dramatically in the

United States in the past few decades and emerging evidence suggests that it might have shifted

from being normal good to an inferior good (Cheng and Kenkel 2010; Kenkel et al. 2014). As

shown in Table 1, about 23 percent of our sample reported being a current smoker and 18 percent

reported being a daily smoker.

6 These measures closely follow those used by Ruhm (2005).

12""

Our next set of outcomes is related to alcohol consumption. Specifically, we examine

three measures of alcohol use: being a current drinker, binge drinking, and chronic drinking. The

outcome of being a “current drinker” measures overall alcohol use, with any reported use over

the past month resulting in a value of 1. Being a “binge drinker” is measured by an indicator,

which takes on the value of 1 if the respondent drank more than five servings of alcohol in one

sitting during the previous month, and 0 otherwise. Lastly, the “chronic drinking” is defined as a

binary indicator for drinking more than 60 alcoholic beverages per month. The evidence on

alcohol use during the economic cycle is also mixed. A number of studies found pro-cyclical

effects (Ruhm and Black 2002, Johansson et al. 2006) while others have found no relationship

(Ogwang and Cho 2009) or even an increase in alcohol use during economic turmoil (Cotti et al.

2013). While the evidence on overall alcohol use is mixed, it has been suggested that heavy and

light drinkers might exhibit an opposite response to poor economic conditions, with heavy

drinkers reducing consumption and light drinkers slightly increasing consumption (Ruhm and

Black 2002). On average, 59 percent of our sample report drinking at least alcoholic beverage in

a month. The prevalence of binge and chronic drinkers are about 16 and 6 percent, respectively.

In addition to changes in consumption patterns of potentially harmful products, the

economic cycle can also affect physical activity patterns and dieting. We examine this potential

channel through the “Physical Activity” and “Obesity” outcomes. In the BRFSS, the respondents

are asked: “During the past month, other than your regular job, did you participate in any

physical activities or exercises such as running, calisthenics, golf, gardening, or walking for

exercise?” If the respondent reported any physical activity in the past 30 days other than that

which he or she does while working, we defined a “Physical Activity” indicator that takes on the

value of 1, and otherwise 0. Similarly, the “Overweight,” “Obese,” and “Severely Obese”

13""

outcomes are indicator variables equal to 1 if the respondent’s self-reported height and weight

result in a Body Mass Index greater than 25, 30, and 35, respectively.7 As shown in Table 1,

approximately 76 percent of sample respondents reported having engaged in physical activity,

while the proportion of the sample who is overweight, obese, and severely obese is 60, 24, and 9

percent, respectively.

The outcome variables thus far have measured potential channels through which the

business cycle can affect health. Our next set of outcomes measures the respondent’s reported

health directly. The outcome of “General Health” captures the respondent’s overall general

health at the time of the interview. While the original response is a categorical variable on a

Likert scale from 1 to 5, we condense respondent’s answers to a binary variable equal to 1 if the

respondent is in very good or excellent health, and 0 otherwise. Likewise, the outcomes of

“Excellent Health” and “Poor Health” are indicators if the respondent reports being in excellent

or poor health, respectively. We also have an outcome to represent those individuals who report

being in either “poor” or “fair” health. The next two outcomes focus on the respondent’s mental

well-being, which includes stress, depression, and problems with emotions over the 30 days

preceding the interview. These outcomes are measured by indicator variables equal to 1 if the

respondent reports poor mental health for more than 10 or 20 days in the past month, and 0

otherwise. About 87 percent of our respondents were in either good or excellent health. The

proportions of our sample in excellent and poor health were 24 percent and 3 percent,

respectively. Finally, approximately 14 percent of our sample reported having mental health

problems for at least 10 days in the past 30 days, while eight percent reported having such

problems for at least 10 days during that period.

7 Body Mass Index is calculated as the ratio of weight in kilograms and height in meters squared.

14""

Finally, we also supplement our analysis with a set of explanatory variables including

binary indicators representing age, gender, race and ethnicity, marital status, and education.

IV. Empirical Method

We estimate a series of regressions that relate changes in health and health behaviors to

macroeconomic conditions along with a vector of individual level characteristics. Specifically,

our basic empirical model is in the following form:

!"#$% = '( + '*+#$% + ,"#$%'- + .# + /$ + 0% + 1!23 +14"#$%1, (1)

where Hismy is one of our outcome measures for individual i living in state s interviewed in

month m of year y. The vector Xismy represents exogenous individual characteristics displayed

in Table 1. The variable of interest in equation (1) is Esmy, one of the two measures of state level

macroeconomic conditions.

In equation (1), we also control for state fixed effects, µs, which would account for

permanent differences across states that may affect health and health behaviors, such as lifestyles

associated with weather patterns, persistent smoking propensities, and state infrastructures on

health care and education. Therefore, the identification of α1 in equation (1) comes from within

state variation in economic conditions over time, rather than fluctuations across states. The δm is

a vector of month fixed effects, which accounts for the impact of seasonality that may exist in

some of the health behaviors such as physical activity (Ruhm 2005). We also control for year

fixed effects, λy, which would capture nationwide trends and shocks that may influence health

behaviors, such as national fluctuations in food and cigarette prices, calorie content in national

chain restaurants, the reduction in payroll tax in 2010, and federal regulations related to health.

We further control for confounding factors that may trend linearly by adding a vector of state-

specific linear time trends denoted by !23. Adding state-specific linear time trends help us

15""

account for unobserved factors that vary within states over time, such as social norms related to

health behaviors like smoking and exercise. These trends also help us control for other state level

time-varying factors such as changes in health care delivery services that closely follow tax

revenues. The εismy is an idiosyncratic random error term. We estimate linear probability models

using Ordinary Least Squares (OLS) and report robust standard errors clustered at the state level.

All the regressions are weighted using the BRFSS sampling weights.8

Estimating equation (1) would shed light into the relationship between macroeconomic

conditions and the outcome variables, but it would not provide any insights into our main

question, i.e., whether this relationship has changed over time with a special emphasis on the

period enveloping the Great Recession. We implement two methods to get at this question. First,

we augment equation (1) by adding a binary variable indicating whether the observation belongs

to a period that coincides with the Great Recession. This specification is similar to the one

estimated by Ruhm (2005), which examines the impact of severe national recessions along with

the effect of unemployment rate. In this specification, the estimate on the Great Recession

indicator reveals whether the recession had any impacts on health and health behaviors beyond

those expected due to the associated higher rates of unemployment (or lower rates of

employment). We also go beyond this specification by estimating regressions with a binary

indicator for Great Recession as well as an interaction between this indicator and each of our

measure of macroeconomic conditions. In these specifications the year fixed effects, 0%, must be

excluded in order to avoid perfect collinearity. Thus, these estimates should be viewed with

caution due to potential omitted variable bias.9

8 Note that we also estimate unweighted regressions. These results are similar to those presented here and are available from the authors upon request. 9 Note that the beginning and end dates for state-level recessions are harder to define and we are aware of no reliable source to gather that information accurately. While the depth and the duration of the recession vary considerably

16""

Second, we turn our attention to our second and also the main strategy, in which we

estimate equation (1) for different periods and observe how the predicted impact of

macroeconomic conditions evolve over time. Following Ruhm (2015ab), we implement this

analysis using differing starting dates and varying lengths of analysis periods. Furthermore, we

also estimate models in which we keep the analysis period fixed, but sequentially move the

sample window to cover all years between 1990-2014 using sample window periods of 5-year,

10-year, and 15-year. These analyses allow us to see how the pattern in the relationship between

macroeconomic conditions and health and health behaviors has evolved over time and between

periods that occurred prior to, during, and after the Great Recession.

V. Results

Table 2 presents the estimates of the relationship between state unemployment rate and

health and health behaviors from equation (1). The first column corresponds to estimates from a

specification with state, month, and year fixed effects along with time-variant state

characteristics. The second column adds state-specific linear time trends to the list of controls in

column 1. Robust standard errors clustered at state level are shown in parentheses.10

The emerging pattern from Table 2 is that the coefficients are mostly small in magnitude

and imprecisely estimated, and this pattern is largely consistent between columns 1 and 2.

Focusing on the estimates in column 2, in which we account for unobserved differences across

states that trend linearly, the only coefficients that are estimated with precision are those of the

two smoking variables and both of them correspond to relatively small effect sizes. In particular,

across states, we adopt the official dating scheme determined by the National Bureau of Economic Research in order to avoid arbitrariness. Our results are largely robust to alternative dating schemes. Nevertheless, the potential discrepancy between the timing of the Great Recession experienced nationwide and state-specific recessions experienced around the same period. 10 We symbolize statistical significance at the 95 and 99 percent levels of confidence. The estimates on other control variables are consistent with those found in the relevant literature are available from the authors upon request.

17""

a one-percentage point increase in the state unemployment rate is associated with a 0.469

percentage-point decrease in the likelihood of being a current smoker and a 0.353 percentage-

point decrease in the likelihood of being a daily smoker. These point estimates translate into an

effect size of approximately 2 percent for both outcomes. The negative relationship between

unemployment and smoking behavior obtained here is largely consistent with the previous

literature (e.g., Ruhm, 2005). The estimates on the other variables draw a mixed picture in terms

of the direction of the relation. For example, chronic drinking appears to be negatively associated

with unemployment while being a current drinker or a binge drinker is positively associated.

Similarly, the relationship between unemployment rate and being overweight appears to be

positive, but it is the opposite for being obese or severely obese. Physical activity has a positive

coefficient, suggesting that a rise in unemployment increases physical exercise. The point

estimate translates into an effect size of 0.431 percentage point or less than one percent

calculated at the sample mean of 0.763.

Regarding the self-reported health outcomes, the evidence in Table 2 is again mixed in

terms of the direction of estimates. For example, unemployment rate appears to be negatively

associated with both being excellent health and fair to poor health, but the estimates are

economically and statistically insignificant in both cases. The pattern is similar for the two

mental health outcomes, where the estimates are with opposite sign, but neither is significant nor

sizeable. Note that the physical and mental health measures in BRFSS are self-assessed and

therefore subject to reporting bias. For example, the response patterns to these questions may be

influenced by economic conditions. This reporting bias may be even more severe if the

responses of survey participants are also influenced by the physical and mental health of other

people close to them. These potential measurement problems may be one reason for the

18""

inconsistency between the findings on health outcomes in our analysis and the procyclical

mortality obtained in the literature. Overall, nine of the 15 estimates, including the two

statistically significant ones, imply that an increase in unemployment is associated with an

improvement in health and health behaviors. Having said that, they are all small in magnitude to

have any meaningful implications for any practical purposes.

The results from the estimation of equation (1) with state employment rate are presented

in Table 3. Similar to Table 2, the estimates are presented from specifications without and with

state-specific linear as shown in columns 1 and 2. Focusing on column 2, the estimates are again

small in magnitude and, with the exception of the two smoking outcomes, statistically

insignificant, suggesting that the relationship between state employment rate and health and

health behaviors is largely inconsequential once observable differences across individuals and

states are accounted for. The estimates on smoking variables are consistent with those in Table 2,

indicating that employment rate and the prevalence of smoking appears to be positively related

with each other. Although none of the estimates on other health behavior outcomes are

significant and sizeable, they appear to point to a pattern of a negative relationship between

employment and prevalence of healthy behaviors, i.e., an increase in state employment rate is

likely to increase the likelihood of obesity as well as being severely obese, and being a chronic

drinker, but likely to decrease engaging in physical activity. While this is consistent with the

previous research in general, it is important to stress that all of the coefficients in Table 3

including those of physical and mental health are small and imprecisely estimated.

One possible interpretation of the results presented in Tables 2 and 3 is that the counter-

cyclical pattern in health and health behaviors obtained by much of the previous research has

become inconsequential when the analysis period is extended to more recent years including the

19""

Great Recession. In order to investigate this possibility further, we next turn our attention to

results from models which include a separate indicator for the Great Recession. In the first set of

results presented in Tables 4 and 5, we report the estimates on our measures of macroeconomic

condition and a binary indicator for the period of Great Recession. As shown in Table 4, the

estimates on unemployment rate are very similar to those displayed in Table 2. The coefficients

on the Great Recession indicator shown in column 3 reveal no association between our outcomes

and the recession. In other words, there are no changes in health and health behaviors during the

period of Great Recession that are beyond those captured by an increase in the state

unemployment rate. All of the estimates in Table 4 are very small ranging from 0.1 percentage

point to 0.6 percentage points, but always imprecisely estimated. The employment rate results

shown in Table 5 point to a similar a story. The only statistically significant estimate is for the

current drinking model, in which the estimated Great Recession effect is equivalent to a

reduction in being current drinker by 0.7 percentage point. However, this effect size is very small

for all practical purposes.

Note that our question of main interest is whether the relationship between indicators of

macroeconomic conditions and health and health behaviors exhibited a different pattern during

the period enveloping the Great Recession. The specification in Tables 4 and 5 show whether the

trajectory of outcomes in health and health behaviors considered in this paper changed in this

period, but do not really address the question of main interest. In order to shed light into this

question, we need to specify an equation that not only contains an indicator for Great Recession,

but also an interaction term between this indicator and our measures of macroeconomic

conditions. The estimates from this specification are shown in Tables 6 and 7 for the

unemployment rate and the employment rate, respectively. The unemployment rate estimates

20""

shown in column 1 are largely consistent with the results from the basic model presented in

Table 2. The unemployment rate is negatively associated with both smoking models and the

estimates are statistically significant at the five percent level in both cases. Regarding the

estimates for the other outcome variables, none are statistically significant, but the overall pattern

appears to favor the notion that an increased unemployment is more likely to result in an

improvement in health and health behaviors than the opposite. For example, the estimates on

chronic drinking, being obese and severely obese, physical exercise, being in fair to poor health

and having at least 10 days of poor mental health are all negative, implying that an increase in

unemployment is associated with an improvement in these outcomes. Exceptions to this pattern

are binge drinking, being overweight, being in at least good health, and experiencing poor mental

health for more than 20 days. But it is important to keep in mind that, if anything, the overall

evidence in Table 6 is weak and therefore should be interpreted with caution since none of the

estimates other than the two that are related to smoking behavior are statistically significant. The

estimates on the Great Recession indicator shown in column 3 are again small and imprecise,

offering no evidence of a particular recession effect beyond that of anticipated by

unemployment.

Next we turn our attention to the interaction coefficients shown in column 5 of Table 6

where we present the estimates on the interaction between the unemployment rate and the Great

Recession. First of all, with the exception of four coefficients, all of the estimates on the

interaction variables are opposite in sign to those of the unemployment rate variable. This

suggests that whatever is the pattern in the relationship between unemployment and health and

health behaviors, that pattern is likely to have been mitigated during the period of Great

Recession for those outcomes. Moreover, of the fifteen outcome variables considered in the

21""

regressions, 11 of them indicate that the unemployment had an effect in the direction of

worsening health and health behaviors during the period coincided with the Great Recession. For

example, the interaction coefficient is positive for both smoking variables, implying that any

smoking reducing effect of unemployment obtained previously appeared to be mitigated during

the Great Recession. A similar dampening effect is also obtained for the outcomes of all three

obesity measures, physical activity, being in good and excellent general health, being in fair to

poor health, and having more than 10 days of poor mental health. Note that the estimate on fair

to poor health is the only one estimated with statistical significance in this group. Among the

outcomes exception to this pattern, the only one estimated precisely is being in poor health,

which suggest that an increase in unemployment reduced the likelihood of being in poor health

during the Great Recession. Although there is no clear explanation for this result, it is

noteworthy that the estimate on unemployment for this model shown in column 1 is essentially

zero. It is important to keep in mind the potential measurement error in this variable for the

reasons mentioned above. This is also the variable with the lowest mean among all the fifteen

outcomes, which may make the regression results more sensitive to outliers.

The corresponding estimates for the interaction between the employment rate and the

Great Recession are shown in Table 7. Consistent with the results shown in Table 6, the

estimates in columns 1 and 5 of Table 7 are discordant in sign with each other in vast majority of

cases – 13 out of 15 to be exact. Keeping in mind that most of the estimates continue to remain

statistically insignificant, one interpretation of this evidence is that the underlying mechanisms

driving the relationship between macroeconomic conditions and health and health behaviors

likely worked in different directions during the period of Great Recession. Focusing on the

estimates on the interaction terms alone, about half of them point to a positive relationship

22""

between the employment rate and health and health behaviors. Moreover, all of the estimates

remain too small to have any meaningful implications.

The overall evidence from Tables 6 and 7 lend some, albeit not strong, support to the

notion that a deterioration in macroeconomic conditions as measured by either rising

unemployment or decreasing employment is associated with an improvement in health and

health behaviors. However, as mentioned earlier, these specifications exclude year fixed effects

to avoid perfect collinearity, raising the possibility of omitted variable bias. Thus, we next turn

attention to our main set of analyses, which is based on estimating the relationship between

macroeconomic conditions and outcomes of interest for different sub-periods, and then exploring

whether there are any detectable changes in this relationship in sub-periods that envelops the

Great Recession. For ease of exposition, we present the results from these analyses in a series of

graphs.

Figure 1A shows the unemployment rate coefficients from regressions in which the

analysis period begins in 1990 and ends in years ranging between 1999 and 2014. Similarly, the

employment rate coefficients are displayed in Figure 1B. Solid lines represent point estimates

and dotted lines show the 95% confidence intervals. It is not surprising that the confidence

intervals become narrower over time since the sample size grows as the analysis period is

extended. Interestingly however, the estimates get closer to zero not only statistically but also

economically in the process and this pattern is present in both Figures 1A and 1B. In other

words, the relationship between macroeconomic conditions and health and health behaviors

appear to become monotonically smaller over time between 1990 and 2014 regardless of the

outcome. Furthermore, the decline in the macroeconomic effect appears to have started in the

years prior to the beginning of the Great Recession and remained largely unchanged in the years

23""

following the recession.11 In addition to this pattern, the estimates have become less volatile over

time as they approach closer to zero.

Next, we specify analysis periods of fixed durations and then sequentially estimate

models for all alternative sample windows permitted by out data. This analysis is performed

using sample window periods of 5-years, 10-years, and 15-years for each of the two measures of

macroeconomic conditions. Figures 2A and 2B display the point estimates along with the 95%

confidence intervals using 5-year sample windows beginning in the specified year for

unemployment and employment rates, respectively. The pattern in these figures imply that

although the effects on health and health behaviors fluctuates around zero throughout the period

between 1990 and 2014, they steadily diminish and approach to zero over time.

The results using 10-year sample windows shown in Figures 3A and 3B are highly

consistent with the pattern observed in the 5-year window analysis, i.e., the relationship between

macroeconomic conditions and health and health behaviors becomes weaker and approaches to

zero over time for both unemployment and employment rates. The seemingly positive (negative)

association between unemployment (employment) for the smoking indicators, overweight and

obesity, physical exercise, general health, and fair to poor health indicators observed for early

periods of windows disappear rather quickly and fluctuate around zero until getting much closer

to zero during the period of Great Recession and afterwards. The relationship is opposite for

binge drinking and the two mental health indicators, but follows the same pattern over time as

the point estimates diminish and become zero. Note that the coefficients are also less volatile for

the 10-year sample windows than the 5-year windows, though we almost always fail to reject the

null hypothesis of no macroeconomic effect on health and health behaviors. 11 One important policy change in the years immediately following the Great Recession is the passage of the Affordable Care Act (ACA) in 2010. One implication of this finding is that the overall pattern of the relationship obtained in our analysis is unlikely to be influenced by the ACA.

24""

Finally, the Figures 4A and 4B display the estimates from sample windows specified at

15-year intervals. The estimates become even smoother in these figures compared to those from

5 and 10-year windows, but they are still imprecisely estimated despite a considerable increase in

sample size. This is consistent with Ruhm (2015a) who concludes from his analyses of the

relationship between unemployment and mortality using similar methods is that the estimates

could be too noisy to be informative when analysis periods are too short.

The results presented so far assume that the relationship between macroeconomic

conditions and health and health behaviors is identical for all demographic groups. Such an

assumption may obscure potentially important differences in the cyclicality of the relationship

across different population groups. We investigated this issue by estimating our models by

gender as well as race and ethnicity. In the interest of space, we only display figures for the

estimates for unemployment rate for the models in which the analysis period begins in 1990 and

ends in years ranging between 1999 and 2014.12 The results for males and females are shown in

Appendix Figures 1A and 1B, and the results for whites, blacks, and Hispanics are displayed in

Figures 2A, 2B, and 2C, respectively. The overall pattern obtained in these regressions is

strongly consistent with the one from the full sample discussed above. One of the consistent

patterns found in the estimations with full sample was the negative relationship between

unemployment rate and two smoking indicators and the positive relationship between

unemployment rate and the likelihood of physical exercise. Appendix Figures 1A and 1B show

that this pattern is largely present among males and not so much among females. Focusing on

race and ethnicity results shown in Appendix Figures 2A-2C, the similar pattern is present for the

two smoking outcomes among whites and Hispanics, but not for blacks. Furthermore, the

12 The results from other models are similar and available from the authors upon request.

25""

relationship appears to be reversed for physical exercise among Hispanics with no clear pattern

among whites and blacks. However, all of these arguments should be viewed with caution since

the vast majority of coefficients are imprecisely estimated. Furthermore, regardless of these

potential differences observed in earlier years, the relationship weakens over time and converge

towards zero during the period of the Great Recession and in the years afterwards.

V. Conclusion

While the effects of economic conditions on health and health behaviors have been well-

studied, the consequences of the Great Recession on these outcomes are not well-understood due

to relative paucity of research focusing on this period. Evidence obtained from studies of

previous recessions may not be applicable to understanding the effect of the recent recession on

health outcomes and health behaviors because the labor market difficulties that individuals

experienced during the recent recession were much more severe in this downturn. In this paper,

we examine the impacts of macroeconomic conditions as measured by state unemployment and

employment rates on a large set of health behaviors and health outcomes using data from BRFSS

between 1990 and 2014 with a close attention to the period enveloping the Great Recession.

Our results provide some support for the notion that weaker macroeconomic conditions

are positively associated with health and health behaviors, although the evidence is stronger for

some of the outcomes (e.g., smoking and physical exercise) than others and is not present for

some of the other outcomes (e.g., experiencing poor mental health). But overall, the estimates are

too small to have any meaningful implications. Furthermore, any evidence for the counter-

cyclical pattern for health and heath behaviors has become weaker over time and largely

disappear in the period prior to the Great Recession and remain near zero in the years since. Our

Finally, the patterns observed for the full sample largely persist when we split our sample by

26""

gender as well as race and gender. The mechanisms for the potential weakening in the

relationship between macroeconomic conditions and health and health behaviors are not well-

understood and remains an important question for future research.

27""

References Ásgeirsdóttir, T., Corman, H., Noonan, K., Ólafsdóttir, Þ., Reichman, N. (2014). "Was the Economic Crisis of 2008 Good for Icelanders? Impact on Health Behaviors." Economics and Human Biology 13: 1– 19. Bobak, M., et al. (2000). Poverty and smoking. Tobacco Control in Developing Countries. F. Chaloupka and P. Jha. Oxford, Oxford University Press. Böckerman, P., et al. (2007). "Does a slump really make you thinner? Finnish micro6level evidence 1978–2002." Health Economics 16(1): 103-107. Business Cycle Dating Committee (2010). Committee Report - 9-20-2010. Catalano, R., et al. (2011). "The health effects of economic decline." Annual review of public health 32: 431-450. Cawley, J., et al. (2011). The impact of the macroeconomy on health insurance coverage: Evidence from the great recession, National Bureau of Economic Research. Centers for Medicaid & Medicare Services (2016). National Health Expensitures 2014 Highlights. https://www.cms.gov/Research-Statistics-Data-and-Systems/Statistics-Trends-and-Reports/NationalHealthExpendData/Downloads/highlights.pdf. Charles, K. K. and P. DeCicca (2008). "Local labor market fluctuations and health: Is there a connection and for whom?" Journal of Health Economics 27(6): 1532-1550. Cheng, K.-W. and D. S. Kenkel (2010). "US cigarette demand: 1944-2004." The BE Journal of Economic Analysis & Policy 10(1). Colman, G. and Dhaval Dave (2013). "Exercise, physical activity, and exertion over the business cycle," Social Science & Medicine, vol. 93(C), pages 11-20 Colman, G. and Dhaval Dave (2014). "Unemployment and Health Behaviors Over the Business Cycle: A Longitudinal View." National Bureau of Economic Research Working Paper No. 20748. Cotti, C., et al. (2013). The Dow is Killing Me: Risky Health Behaviors and the Stock Market. Available at SSRN. Currie, J. and E. Tekin (2015). Is there a Link Between Foreclosure and Health? American Economic Journal: Economic Policy 7(1): 63-94. Dee, T. S. (2001). "Alcohol abuse and economic conditions: evidence from repeated cross6sections of individual6level data." Health Economics 10(3): 257-270.

28""

Dehejia, R. and A. Lleras-Muney (2004). "Booms, busts, and babies' health." The Quarterly Journal of Economics 119(3): 1091-1130. Ettner, S. L. (1997). "Measuring the human cost of a weak economy: Does unemployment lead to alcohol abuse?" Social Science & Medicine 44(2): 251-260. Gerdtham, U.-G. and C. J. Ruhm (2006). "Deaths rise in good economic times: evidence from the OECD." Economics & Human Biology 4(3): 298-316. Granados, J. A. T. (2005). "Recessions and mortality in Spain, 1980–1997." European Journal of Population/Revue européenne de Démographie 21(4): 393-422. Holahan, J. (2011). "The 2007–09 recession and health insurance coverage." Health Affairs 30(1): 145-152. Johansson, E., et al. (2006). "Alcohol-related mortality, drinking behavior, and business cycles." The European Journal of Health Economics 7(3): 212-217. Kenkel D.S. & Maximilian D. Schmeiser & Carly Urban (2014). "Is Smoking Inferior?: Evidence from Variation in the Earned Income Tax Credit," Journal of Human Resources, vol. 49(4), pages 1094-1120. Kochhar, R. (2011). Two Years of Economic Recovery: Women Lose Jobs, Men Find Them Pew Research Social & Demographic Trends Project. Washington DC. Macy, J. T., et al. (2013). "Predictors of health behaviors after the economic downturn: A longitudinal study." Social Science & Medicine. McInerney, M. and J. M. Mellor (2012). "Recessions and Seniors’ Health; Health Behaviors; and Healthcare Use: Analysis of the Medicare Current Beneficiary Survey." Journal of Health Economics. Mortgage Bankers Association (2010). "Delinquencies and Foreclosure Starts Decrease in Latest MBA National Delinquency Survey." from http://www.mbaa.org/NewsandMedia/PressCenter/73799.htm. Neumayer, E. (2004). "Recessions lower (some) mortality rates:: evidence from Germany." Social Science & Medicine 58(6): 1037-1047. Ogwang, T. and D. I. Cho (2009). "Economic determinants of the consumption of alcoholic beverages in Canada: a panel data analysis." Empirical Economics 37(3): 599-613. Ruhm, C. J. (1995). "Economic conditions and alcohol problems." Journal of Health Economics 14(5): 583-603.

29""

Ruhm, C. J. (2000). "Are recessions good for your health?" The Quarterly Journal of Economics 115(2): 617-650. Ruhm, C. J. (2003). "Good times make you sick." Journal of Health Economics 22(4): 637-658. Ruhm, C. J. (2005). "Healthy living in hard times." Journal of Health Economics 24(2): 341-363. Ruhm, C. J. (2007). "A healthy economy can break your heart." Demography 44(4): 829-848. Ruhm, C. J. (2015a). Recessions, Healthy No More? Journal of Health Economics 42:17-28. Ruhm, C. J. (2015b). Health Effects of Economic Crises. National Bureau of Economic Research Working Paper No. 21604. Ruhm, C. J. and W. E. Black (2002). "Does drinking really decrease in bad times?" Journal of Health Economics 21(4): 659-678. U.S. Census Bureau (2012). Income, Poverty, and Health Insurance Coverage in the United States: 2911. https://www.census.gov/newsroom/releases/archives/income_wealth/cb12-172.html U.S. Department of Agriculture (2012). Building a Healthy America: A Profile of the Supplemental Nutrition Assistance Program. Food and Nutrition Service, Office of Research and Analysis. http://www.fns.usda.gov/sites/default/files/BuildingHealthyAmerica.pdf. White, C. and J. D. Reschovsky (2012). Great Recession Accelerated Long-Term Decline of Employer Health Coverage NIHCR. Research Brief No. 8. Willard, R., et al. (2012). "Impact of the 2008–2010 economic recession on local health departments." Journal of Public Health Management and Practice 18(2): 106-114. Xu, X. (2013). "The business cycle and health behaviors." Social Science & Medicine Volume 77, 126–136.

30##

Figure 1A: Changes in the Relationship Between Unemployment and Health and Health Behaviors over Time: Sample Begins in 1990 and Continues through the Specified Year

Note: The dotted lines represent the 95% confidence intervals.

31##

Figure 1B: Changes in the Relationship Between Employment and Health and Health Behaviors over Time: Sample Begins in 1990 and Continues through the Specified Year


32##

Figure 2A: Changes in the Relationship Between Unemployment and Health and Health Behaviors over Time: 5-Year Windows


33##

Figure 2B: Changes in the Relationship Between Employment and Health and Health Behaviors over Time: 5-Year Windows


34##



35##



36##



37##



38##

Table 1: Descriptive Statistics Variable Mean Standard Deviation Outcomes

Current Smoker 0.230 0.421 Daily Smoker 0177 0.359 Current Drinker 0.587 0.486 Binge Drinker 0.156 0.401 Chronic Drinker 0.057 0.233 Physical Exercise 0.763 0.410 Overweight 0.599 0.472 Obese 0.240 0.452 Severely Obese 0.087 0.282 General Health 0.874 0.305 Excellent Health 0.240 0.420 Poor Health 0.029 0.121 Fair to Poor Health 0.126 0.331 Mental10: Poor Mental Health>10 days/month 0.142 0.333 Mental20: Poor Mental Health>20 days/month 0.078 0.248 Control Variables State Employment Rate 0.613 0.040 State Unemployment Rate 0.063 0.024 Age: 25-29 0.164 0.352 Age: 30-34 0.185 0.385 Age: 35-39 0.170 0.367 Age: 40-44 0.177 0.388 Age: 45-49 0.153 0.367 Age: 50-54 0.151 0.375 Male 0.484 0.498 Married 0.638 0.479 Divorced 0.104 0.295 Widowed 0.012 0.098 Other Marital Status a 0.247 0.435 Less Than High School a 0.105 0.291 High School 0.263 0.434 Some College Education 0.273 0.443 College 0.358 0.487 White 0.664 0.475 Black 0.117 0.318 Other a 0.109 0.264 Hispanic 0.110 0.362

Note: Data are from1990 to 2014 years of the BRFSS. Number of observations are 2,362,954 for the summary statistics. a Omitted category.

39##

Table 2: Estimates of the Relationship between Unemployment and Health and Health Behaviors Outcome I II Current Smoker -0.382** -0.469**

(0.120) (0.136)

Daily Smoker -0.321** -0.353**

(0.085) (0.101)

Current Drinker 0.083 0.026

(0.179) (0.160)

Binge Drinker 0.126 0.026

(0.145) (0.083)

Chronic Drinker -0.060 -0.092

(0.040) (0.057)

Overweight -0.017 0.019

(0.077) (0.076)

Obese -0.009 -0.043

(0.104) (0.089)

Severely Obese -0.012 -0.019

(0.063) (0.039)

Physical Activity 0.507* 0.431

(0.240) (0.246)

General Health -0.108 0.062

(0.119) (0.069)

Excellent Health -0.052 -0.012

(0.098) (0.072)

Poor Health 0.045 0.024

(0.027) (0.031)

Fair to Poor Health 0.108 -0.062

(0.119) (0.069)

Mental Health Problems >10 days 0.100 -0.016

(0.083) (0.092)

Mental Health Problems >20 days 0.119* 0.047

(0.056) (0.054)

Month Fixed Effects Yes Yes Year Fixed Effects Yes Yes State Fixed Effects Yes Yes State-specific Linear Trends No Yes

Notes: Standard errors are clustered at state. A * and ** indicate that the estimate is statistically significance at the 95 and 99 percent levels of confidence, respectively.

40##

Table 3: Estimates of the Relationship between Employment and Health and Health Behaviors Outcome I II Current Smoker 0.123 0.306*

(0.087) (0.128)

Daily Smoker 0.118 0.277*

(0.073) (0.106)

Current Drinker -0.088 0.078

(0.127) (0.161)

Binge Drinker -0.122 -0.013

(0.142) (0.100)

Chronic Drinker 0.059 0.088

(0.058) (0.071)

Overweight 0.001 0.024

(0.060) (0.074)

Obese -0.031 0.086

(0.075) (0.066)

Severely Obese -0.042 0.018

(0.048) (0.031)

Physical Activity -0.173 -0.024

(0.169) (0.180)

General Health 0.078 -0.007

(0.042) (0.041)

Excellent Health 0.034 0.045

(0.072) (0.073)

Poor Health -0.028 -0.012

(0.022) (0.021)

Fair to Poor Health -0.078 0.007

(0.042) (0.041)

Mental Health Problems >10 days -0.125* -0.048

(0.060) (0.061)

Mental Health Problems >20 days -0.119* -0.074

(0.045) (0.045)

Month Fixed Effects Yes Yes Year Fixed Effects Yes Yes State Fixed Effects Yes Yes State-specific Linear Trends No Yes

Notes: Standard errors are clustered at state. A * and ** indicate that the estimate is statistically significance at the 95 and 99 percent levels of confidence, respectively.

41##

Table 4: Estimates of the Relationship Between Unemployment and Health and Health Behaviors with an Indicator for Great Recession Outcome Unemployment Rate Great Recession Current Smoker -0.474** (0.138) -0.002 (0.002) Daily Smoker -0.356** (0.108) -0.001 (0.003) Current Drinker 0.007 (0.166) -0.006 (0.003) Binge Drinker 0.016 (0.086) -0.003 (0.004) Chronic Drinker -0.096 (0.057) -0.001 (0.002) Overweight 0.008 (0.081) -0.004 (0.005) Obese -0.027 (0.086) 0.006 (0.004) Severely Obese -0.008 (0.040) 0.004 (0.003) Physical Activity 0.445 (0.257) 0.005 (0.005) General Health 0.056 (0.077) -0.002 (0.003) Excellent Health -0.015 (0.078) -0.001 (0.004) Poor Health 0.018 (0.031) -0.002 (0.001) Fair to Poor Health -0.056 (0.077) 0.002 (0.003) Mental Health Problems >10 days -0.011 (0.094) 0.001 (0.003) Mental Health Problems >20 days 0.051 (0.054) 0.001 (0.002)

Notes: Standard errors are clustered at state. A * and ** indicate that the estimate is statistically significance at the 95 and 99 percent levels of confidence, respectively. In addition to time-variant state characteristics, the models also control for state and month fixed effects, and state specific linear time trends.

42##

Table 5: Estimates of the Relationship Between Employment and Health and Health Behaviors with an Indicator for Great Recession Outcome Employment Rate Great Recession Current Smoker 0.308* (0.130) -0.000 (0.002) Daily Smoker 0.279* (0.111) -0.001 (0.003) Current Drinker 0.092 (0.165) -0.007* (0.003) Binge Drinker -0.007 (0.104) -0.003 (0.004) Chronic Drinker 0.097 (0.071) -0.001 (0.002) Overweight 0.032 (0.078) -0.005 (0.005) Obese 0.077 (0.065) 0.005 (0.004) Severely Obese 0.008 (0.028) 0.004 (0.003) Physical Activity -0.027 (0.186) 0.001 (0.005) General Health 0.001 (0.045) -0.002 (0.003) Excellent Health 0.047 (0.078) -0.001 (0.004) Poor Health -0.006 (0.020) -0.002 (0.001) Fair to Poor Health -0.001 (0.045) 0.002 (0.003) Mental Health Problems >10 days -0.056 (0.064) 0.002 (0.003) Mental Health Problems >20 days -0.079 (0.046) 0.001 (0.002)


43##

Table 6: Estimates of the Relationship Between Unemployment and Health and Health Behaviors with an Indicator for Great Recession and Interaction with Employment Outcome

Unemployment Rate

Great Recession Unemployment Rate* Great Recession

Current Smoker -0.488** (0.143) -0.011 (0.010) 0.126 (0.121) Daily Smoker -0.359** (0.111) -0.003 (0.008) 0.028 (0.084) Current Drinker 0.036 (0.171) 0.013 (0.012) -0.241 (0.142) Binge Drinker 0.027 (0.089) 0.005 (0.006) -0.103 (0.086) Chronic Drinker -0.086 (0.056) 0.006 (0.010) -0.087 (0.128) Overweight 0.002 (0.084) -0.008 (0.010) 0.048 (0.100) Obese -0.033 (0.091) 0.002 (0.009) 0.050 (0.113) Severely Obese -0.015 (0.042) -0.001 (0.008) 0.063 (0.113) Physical Activity 0.459 (0.257) 0.014 (0.011) -0.112 (0.115) General Health 0.074 (0.076) 0.010 (0.006) -0.147* (0.069) Excellent Health 0.003 (0.078) 0.010 (0.008) -0.143 (0.096) Poor Health 0.009 (0.031) -0.008** (0.003) 0.074* (0.028) Fair to Poor Health -0.074 (0.076) -0.010 (0.006) 0.147* (0.069) Mental Health Problems >10 days -0.015 (0.091) -0.001 (0.008) 0.029 (0.094) Mental Health Problems >20 days 0.053 (0.051) 0.002 (0.006) -0.019 (0.066)


44##

Table 7: Estimates of the Relationship Between Employment and Health and Health Behaviors with an Indicator for Great Recession and Interaction with Employment Outcome

Employment Rate

Great Recession Employment Rate* Great Recession

Current Smoker 0.309* (0.131) 0.017 (0.025) -0.029 (0.042) Daily Smoker 0.280* (0.111) 0.005 (0.021) -0.009 (0.035) Current Drinker 0.087 (0.165) -0.066 (0.048) 0.097 (0.080) Binge Drinker -0.010 (0.104) -0.052* (0.024) 0.081* (0.037) Chronic Drinker 0.096 (0.071) -0.016 (0.019) 0.026 (0.032) Overweight 0.033 (0.078) 0.013 (0.029) -0.030 (0.047) Obese 0.080 (0.065) 0.048 (0.031) -0.071 (0.049) Severely Obese 0.008 (0.028) 0.006 (0.024) -0.004 (0.039) Physical Activity -0.030 (0.186) -0.028 (0.035) 0.049 (0.056) General Health 0.002 (0.044) 0.012 (0.018) -0.024 (0.028) Excellent Health 0.046 (0.079) -0.011 (0.042) 0.016 (0.068) Poor Health -0.007 (0.020) -0.013 (0.007) 0.018 (0.011) Fair to Poor Health -0.002 (0.044) -0.012 (0.018) 0.024 (0.028) Mental Health Problems >10 days -0.058 (0.064) -0.021 (0.020) 0.038 (0.033) Mental Health Problems >20 days -0.081 (0.045) -0.024 (0.015) 0.041 (0.023)


45##

-3-2

-10

2000 2005 2010 2015

Current Smoker

-2.5

-2-1

.5-1

-.50

2000 2005 2010 2015

Daily Smoker

-1.5

-1-.5

0.5

2000 2005 2010 2015

Current Drinker

-1-.5

0.5

12000 2005 2010 2015

Binge Drinker

-2-1

01

2

2000 2005 2010 2015

Chronic Drinker

-1.5

-1-.5

0.5

2000 2005 2010 2015

Overweight

-1-.5

0.5

2000 2005 2010 2015

Obese

-.4-.2

0.2

2000 2005 2010 2015

Severely Obese

-.50

.51

1.52

2000 2005 2010 2015

Physical Exercise

-1-.5

0.5

11.5

2000 2005 2010 2015

General Health

-1.5

-1-.5

0.5

1

2000 2005 2010 2015

Excellent Health

-.6-.4

-.20

.2.4

2000 2005 2010 2015

Poor Health

-1.5

-1-.5

0.5

1

2000 2005 2010 2015

Fair to Poor Health

-1-.5

0.5

1

2000 2005 2010 2015

Mental Health Problems>10 days/month

-.50

.5

2000 2005 2010 2015


Appendix Figure 1A: Changes in the Relationship Between Unemployment and Health and Health Behaviors among Males over Time: Sample Begins in 1990 and Continues through the Specified Year


46##

Appendix Figure 1B: Changes in the Relationship Between Unemployment and Health and Health Behaviors among Females over Time: Sample Begins in 1990 and Continues through the Specified Year


-2-1

.5-1

-.50

.5

2000 2005 2010 2015

Current Smoker

-1.5

-1-.5

0.5

2000 2005 2010 2015

Daily Smoker

-.50

.51

1.5

2000 2005 2010 2015

Current Drinker

-.4-.2

0.2

.4.6

2000 2005 2010 2015

Binge Drinker

-.50

.51

2000 2005 2010 2015

Chronic Drinker

-1-.5

0.5

2000 2005 2010 2015

Overweight

-.8-.6

-.4-.2

0.2

2000 2005 2010 2015

Obese

-.3-.2

-.10

.1.2

2000 2005 2010 2015

Severely Obese

-1-.5

0.5

1

2000 2005 2010 2015

Physical Exercise

-1-.5

0.5

1

2000 2005 2010 2015

General Health

-1-.5

0.5

1

2000 2005 2010 2015

Excellent Health

-.4-.2

0.2

.4

2000 2005 2010 2015

Poor Health

-1-.5

0.5

1

2000 2005 2010 2015

Fair to Poor Health

-.50

.51

2000 2005 2010 2015


-.20

.2.4

.6.8

2000 2005 2010 2015


47##

-2.5

-2-1

.5-1

-.50

2000 2005 2010 2015

Current Smoker

-2-1

.5-1

-.50

2000 2005 2010 2015

Daily Smoker

-1-.5

0.5

2000 2005 2010 2015

Current Drinker

-.50

.51

2000 2005 2010 2015

Binge Drinker

-1-.5

0.5

11.

5

2000 2005 2010 2015

Chronic Drinker

-1-.5

0.5

2000 2005 2010 2015

Overweight

-.4-.2

0.2

2000 2005 2010 2015

Obese

-.3-.2

-.10

.1

2000 2005 2010 2015

Severely Obese

-1-.5

0.5

11.

5

2000 2005 2010 2015

Physical Exercise

-.6-.4

-.20

.2.4

2000 2005 2010 2015

General Health

-1-.5

0.5

2000 2005 2010 2015

Excellent Health

-.2-.1

0.1

.2.3

2000 2005 2010 2015

Poor Health

-.4-.2

0.2

.4.6

2000 2005 2010 2015

Fair to Poor Health

-.20

.2.4

.6.8

2000 2005 2010 2015


-.20

.2.4

.6

2000 2005 2010 2015


Appendix Figure 2A: Changes in the Relationship Between Unemployment and Health and Health Behaviors among Whites over Time: Sample Begins in 1990 and Continues through the Specified Year


48##

-2-1

01

2

2000 2005 2010 2015

Current Smoker

-1.5

-1-.5

0.5

1

2000 2005 2010 2015

Daily Smoker

-10

12

2000 2005 2010 2015

Current Drinker

-1-.5

0.5

11.

5

2000 2005 2010 2015

Binge Drinker

-1-.5

0.5

1

2000 2005 2010 2015

Chronic Drinker

-3-2

-10

1

2000 2005 2010 2015

Overweight

-3-2

-10

1

2000 2005 2010 2015

Obese

-1-.5

0.5

1

2000 2005 2010 2015

Severely Obese

-2-1

01

23

2000 2005 2010 2015

Physical Exercise

-2-1

01

2000 2005 2010 2015

General Health

-10

12

34

2000 2005 2010 2015

Excellent Health

-1.5

-1-.5

0.5

2000 2005 2010 2015

Poor Health

-10

12

2000 2005 2010 2015

Fair to Poor Health

-2-1

01

2

2000 2005 2010 2015


-10

12

2000 2005 2010 2015


Appendix Figure 2B: Changes in the Relationship Between Unemployment and Health and Health Behaviors among Blacks over Time: Sample Begins in 1990 and Continues through the Specified Year


49##

-2-1

.5-1

-.50

.5

2000 2005 2010 2015

Current Smoker

-2-1

01

2000 2005 2010 2015

Daily Smoker

-3-2

-10

1

2000 2005 2010 2015

Current Drinker

-1.5

-1-.5

0.5

1

2000 2005 2010 2015

Binge Drinker

-2-1

01

2000 2005 2010 2015

Chronic Drinker

-2-1

01

2

2000 2005 2010 2015

Overweight

-3-2

-10

1

2000 2005 2010 2015

Obese

-1-.5

0.5

2000 2005 2010 2015

Severely Obese

-4-3

-2-1

01

2000 2005 2010 2015

Physical Exercise

-10

12

3

2000 2005 2010 2015

General Health

-3-2

-10

12

2000 2005 2010 2015

Excellent Health

-1-.5

0.5

1

2000 2005 2010 2015

Poor Health

-2-1

01

2000 2005 2010 2015

Fair to Poor Health

-3-2

-10

12

2000 2005 2010 2015


-2-1

01

2

2000 2005 2010 2015


Appendix Figure 2C: Changes in the Relationship Between Unemployment and Health and Health Behaviors among Hispanics over Time: Sample Begins in 1990 and Continues through the Specified Year


Documents

HEALTH AND HEALTH BEHAVIORS DURING THE WORST OF …According to data from the National Health Expenditure Accounts, the health care spending represents a significant and increasing